Streptomycin is the only effective form of chemical control New Zealand growers have to combat fire blight (Erwinia amylovora). As streptomycin is an antibiotic, strains of fire blight could potentially develop resistance to streptomycin, in the same way humans have developed resistance to many of the medical antibiotics. Therefore it is logical that streptomycin is applied only when absolutely necessary. Streptomycin is also an expensive spray so most growers would be keen to minimise its use in their programme for cost reasons, and to conform with the requirements of Integrated Fruit Production (IFP). In an IFP programme, the aim is to produce marketable fruit with fewer fungicide applications that are safe as possible to humans and the environment.

Some seasons in New Zealand, conditions are unfavourable for fire blight development and yet streptomycin is still applied by growers as a preventative measure. If we could predict when a fire blight infection event is likely to occur, we could then decide whether streptomycin is required or not.

Computer models, which predict fire blight infections based on weather data, have been developed in the United States and are now being trialed in New Zealand. To date results look promising under New Zealand conditions.

What is needed for an infection event ?

Infection by any disease requires a susceptible host, disease inoculum and the correct environmental conditions. If we assume that the fire blight bacteria is present and the plant is susceptible, the following events need to occur in sequence for a fire blight infection to take place:

1.	Flowers open with stigmas and petals intact.
2.	The accumulation of at least 110 degree hours* when temperatures
	are greater than 18.3ºC after first bloom.
3.	Wetting caused either by dew or 0.25 mm of rain
	(or at least 2.5 mm of rain the previous day).
4.	An average daily temperature of at least 15.6ºC.

* A degree hour is one hour for every degree centigrade above the threshold temperature (in this case 18.3ºC).

Growers can install a weather station to measure maximum and minimum temperature and rainfall, or use information from an automatic weather datalogger. However, the computation of heat units required for infection and the occurrence of disease symptoms can be complicated. This is why dedicated computer-based systems are so useful.

Computerised fire blight prediction systems

The most successful computerised prediction system to be developed in the United States is the MARYBLYT™ model. MARYBLYT™ has been trialed for over 10 years in the United States and is now being used by American growers and consultants to predict fire blight.

The MARYBLYT™ programme uses the infection criteria described above to identify infection events and predict the development of symptoms. Computer printouts give the daily temperatures, wetness and likelihood of infection which can be interpreted to decide whether a streptomycin spray is required.

Dr Chin Gouk and her team from HortResearch began trialing the MARYBLYT™ model in 1991 to see how it performs under New Zealand conditions.

In 1992, work was also started on a computer based prediction system specifically designed for the New Zealand pipfruit industry. This system uses the criteria from MARYBLYT™ and another model to predict an infection event. It has the advantage that it can be run using automatic weather data input from the Orchard 2000 network of dataloggers.

Trialing the systems in New Zealand

During the 1994/95 season, both the MARYBLYT™ and HortResearch programmes were extensively evaluated on apple and pear orchards in Waikato and Hawke’s Bay.

During the main flowering period the average daily temperatures were too low for infection to occur and, correspondingly, no infection periods were predicted by the computer systems.

From November onwards there were four infection events in Hawke’s Bay and five in the monitored Waikato orchards on Royal Gala and Braeburn trees. All of these were accurately predicted because symptoms of fire blight on the blossoms occurred either on the day, or within a day of the dates they were expected to appear.

Fire blight did not show up on orchards which had reached complete petal fall when the infection events occurred. Royal Gala trees developed fire blight symptoms whereas Braeburn, which had finished flowering, did not. Infection events need to coincide with flowering (including late flushes) for the disease to develop. This will vary between cultivars and orchards.

Two out of the eight Hawke’s Bay orchards monitored did not have any fire blight despite conditions being conducive to fire blight development. This implies that previous disease history should be taken into account when using predictive information - if there is no history of fire blight then the bacterial inoculum may not be present.

The prediction systems were not as accurate on pears, as many of the fire blight symptoms appeared earlier than predicted by the computer. The criteria for number of growing degree days after infection may need to be adjusted to improve the accuracy.

Where to now ?

Both the MARYBLYT™ and the HortResearch predictive models have been shown to accurately predict the onset of fire blight at the blossom stage on Royal Gala. Over the next three years, the models will continue to be tested on different varieties in other regions so we can be sure the programmes are accurate for New Zealand conditions before they are commercialised.

In the meantime, growers should pay attention to weather conditions for the timing of streptomycin sprays and minimise use of this antibiotic whenever possible.

Acknowledgments

HortResearch acknowledges ENZA New Zealand (International), Ministry of Agriculture and Fisheries (Policy Division) and the Foundation for Research, Science and Technology for funding.

References

GOUK, S.C., R.J. BEDFORD, S.O. HUTCHINGS, L. COLE and M.D VOYLE (1996). Evaluation of the MARYBLYT™ model for predicting fire blight blossom infection in New Zealand. Acta Horticulturae. 411: 109-116.

GOUK, S.C., R.J. BOYD and S.O. HUTCHINGS (1996). Applications of the MARYBLYT™ computer model for identifying infection risk for fire blight of apple. Proc N.Z Plant Protection Society Conference. (In press).